Sega757/Fast-AtomRAG-Core
Lightweight, zero-broker hybrid retrieval engine for RAG. Combines vector search, local Personalized PageRank (PPR) over Atomic Entity Graphs (AEG), and Reciprocal Rank Fusion (RRF). Features SLA-driven cost-aware routing and strict SQLite WAL persistence. Built for extreme low-latency and AI agent orchestration.
GitHub repository with 85 stars and 0 forks.
Language: Python
Topics: ai-agents, cost-aware-routing, fast-api, hybrid-search, information-retrieval, knowledge-graph, low-latency, nlp, pagerank, rag